Artificial Intelligence Is Ready for the Legal Industry's Attention | The American Lawyer – Law.com
Some attorneys have a tendency to think of artificial intelligence (AI) as something akin to cellphones in the 1980s or computers in the 1970s. That is to say, expensive enough that only the law firms or legal departments with the deepest pockets can tap into its nebulous potential.
But AI is far from a mysterious or inaccessible innovation. For years, it has powered online chatbots and search engines, voice-activated assistants and video games. And most of these are widely available and affordable. So why shouldn’t it be the same for the legal industry?
Today, AI is the unseen engine behind a host of off-the-shelf legal technology tools. It is so embedded in products from e-discovery to e-billing and contract analytics to legal research that it’s not hard to see a day when AI becomes unavoidable.
“You’re going to see it be a part of any legal tech product … whether there is an initiative to go out and buy or develop AI products or not,” says Jeff Marple, director of innovation for the legal department at insurance company Liberty Mutual.
Even with these advances though, significant barriers to adoption remain. And the ability to create bespoke, proprietary AI systems is still largely the domain of a well-funded few at the upper echelons of the legal market.
Yet there are some signs that even this is shifting. The rise and continued growth of “productized” AI means the technology can be more accessible to small and midsize law firms and legal departments than ever before. AI products will also likely get better and easier to deploy over time, and some legal teams are creatively pushing the limits of what AI can do and be in a market that has traditionally been underserved.
The days in which small and midsize buyers were little more than an afterthought for AI developers, it seems, are long gone.
How the Cost Was Cut
As AI becomes more of an essential component of legal tech platforms, it’s becoming cheaper to use. “The implementation cost is coming down significantly,” Marple notes. There are several reasons behind this growing affordability, including the advancement of a wholly different technology: cloud computing.
In the past, if someone wanted to bring an advanced technology such as AI in-house, they had to support it with physical servers and staff to configure and manage the software—essentially the body and muscles behind the brain. But now, the paradigm has completely shifted, says Ryan Duguid, chief evangelist at workflow automation company Nintex and a tech veteran who previously held various positions at Microsoft.
“All that changed when the cloud became a reality,” he explains. “The delivery of the software got cheaper, faster and better.”
Now, legal teams can try an AI platform without any upfront investment—the body and muscles are handled by the legal tech company off-site. “The beauty of the [cloud-based delivery] model is I can stop my investment at any time and try something else,” Duguid adds.
In addition to using the cloud, AI products are also less costly because they can be deployed out of the box. Many AI systems, especially those reviewing documents or contracts, need to be trained before they can read and identify information with an almost-human level of understanding. Such machine learning is a resource-intensive, time-consuming process. But it is one that legal tech providers will carry out ahead of time, so a legal research platform, for instance, knows how to identify relevant language in complaints and rulings as soon as it’s turned on.
Having a pretrained product is a huge benefit for small and midsize law firms and legal departments, many of which are unlikely to have the type of large data sets needed to train an AI tool in the first place. But that doesn’t mean they can’t help to improve the tool’s accuracy.
Productized AI, after all, is continuously learning and getting smarter with every use, like a child whose reading comprehension expands the more he or she reads. “As it gains wider adoption, the use of the tool makes the tool better,” explains Aaron Crews, chief data analytics officer at Littler Mendelson and former general counsel at AI startup TextIQ.
The Fine Print
Since many of today’s AI products are meant to work with a user base as broad as possible, they are pretrained to read only the most widely applicable data—commonly used contracts clauses, for example, or routine court filings. When it comes to working with unique information, the tools may be less than effective.
Alan Winchester, a member of law firm Harris Beach, says AI tools “have potential” and notes that his firm, which has around 200 attorneys, uses e-discovery and contract review AI products. But even so, he says, “I haven’t seen pretrained technology that is great.” Yet for some, productized AI’s limited scope may be a deal breaker. Nintext senior corporate counsel Camden Hillas says her two-person legal department is currently in the process of evaluating AI tools on the market.
“Especially as a smaller group, you may create some efficiencies by using AI technology,” she says. “However, you can’t guarantee the value just yet because there is too big of a risk of potential mistakes, and a smaller group can’t really afford that.”
To be sure, even if the accuracy of such products increases, which, in theory, they should over time, there’s still the problem of figuring out how to get your money’s worth.
“AI is typically all about value, what can I deliver for a cost,” Liberty Mutual’s Marple says. “And in order to be able to calculate that and see what the changes are by using a product versus not using a product … you need to have good operationalized metrics.”
Setting up processes to quantitatively measure and track return on investment, however, is an added cost that may be more than some can shoulder. Since there is little uniformity with how legal tech providers price similar AI products, it can be a complicated affair.
“In the AI space, everyone is super cagey about what they charge,” Nintext’s Duguid says. He further explains, “A lot of vendors are essentially pulling the value-based pricing scam—‘So I’m not going to put my price on the website because I am going to charge you whatever I think you’re prepared to pay for it’—which can make these evaluations tricky.”
Despite productized AI’s problems, it’s often the only option small- and medium-sized buyers have. Creating their own bespoke AI solution is usually too expensive.
In that scenario, “you’re going to have to hire a bunch of top-end developers, and those developers are going to cost you … $350 or $450 an hour. They will be [working] away a long time on a high-risk project, and you’re going to hope you get a return on it,” Duguid says. But some small and midsize law firms are not ceding the creative ground. While not creating AI solutions as complex as the larger players, they are still showing they can leverage AI technology as developers.
Harris Beach, for instance, used productized AI solutions to support its custom-built regulatory compliance tool CyMetric, which was released at the end of 2018 through its legal tech subsidiary Caetra. The tool drafts policies and controls for companies required to meet certain regulations. “We are using AI to help us identify the regulations and elements of the regulations we need to map … to come up with a model and a security compliance recommendation for customers on the tool,” Winchester explains.
In addition to Harris Beach, Chicago-based Actuate Law, a 13-attorney law firm that launched a tech subsidiary called Quointec in early this year, recently developed the “Data Breach and Privacy Advisor.” The AI-powered tool helps inform clients of their legal obligations after a security incident.
To make the custom tool, Actuate partnered with AI developer Neota Logic, which provided the “building blocks” for the AI engine, and Thomson Reuters Legal Managed Services, which developed the “curated dashboard” front end, Actuate partner and Quointec’s chief innovation evangelist Martin Tully explains.
The tool’s AI is not as advanced as machine-learning software that gets smarter with every training. Instead, it uses a “very fancy decision-tree-type logic to replicate a legal analysis that a human would do,” Tully says. He adds that clients “essentially get the same output they would if they spoke to a human lawyer.”
So how did a small law firm get Neota Logic and Thomson Reuters to collaborate with it on a proprietary tool? A lot of it came down to networking. “Both myself and Martin Tully had relationships with Thomson Reuters Legal Managed Services,” Actuate partner and Quoint ec’s chief innovation strategist Dara Tarkowski says.
Tarkowski adds, however, that you don’t need personal connections to get tech developers’ support. “[The] community of those committed to legal innovation is quite small in my view,” she says, noting that it’s not hard to get in touch with legal tech companies if you’re an innovator.
Of course, in today’s market, one need not be an innovator to be able to leverage AI. “You don’t need to have a tech subsidiary in order to utilize and offer some of these [AI] tools and services that already exist in the marketplace,” Tully says. “But if you want to build new tools and create new solutions for problems using AI technology, then having a tech subsidiary comes in real handy.”
So for small and midsize law firms and legal departments, the opportunity to move beyond being passive buyers may be there for the taking. At the end of the day, where there’s a will, there’s a way.